include/sys/range_tree.h

   1 /*
   2  * CDDL HEADER START
   3  *
   4  * The contents of this file are subject to the terms of the
   5  * Common Development and Distribution License (the "License").
   6  * You may not use this file except in compliance with the License.
   7  *
   8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
   9  * or https://opensource.org/licenses/CDDL-1.0.
  10  * See the License for the specific language governing permissions
  11  * and limitations under the License.
  12  *
  13  * When distributing Covered Code, include this CDDL HEADER in each
  14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
  15  * If applicable, add the following below this CDDL HEADER, with the
  16  * fields enclosed by brackets "[]" replaced with your own identifying
  17  * information: Portions Copyright [yyyy] [name of copyright owner]
  18  *
  19  * CDDL HEADER END
  20  */
  21 /*
  22  * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  23  * Use is subject to license terms.
  24  */
  25
  26 /*
  27  * Copyright (c) 2013, 2019 by Delphix. All rights reserved.
  28  */
  29
  30 #ifndef _SYS_RANGE_TREE_H
  31 #define _SYS_RANGE_TREE_H
  32
  33 #include <sys/btree.h>
  34 #include <sys/dmu.h>
  35
  36 #ifdef  __cplusplus
  37 extern "C" {
  38 #endif
  39
  40 #define RANGE_TREE_HISTOGRAM_SIZE       64
  41
  42 typedef struct range_tree_ops range_tree_ops_t;
  43
  44 typedef enum range_seg_type {
  45         RANGE_SEG32,
  46         RANGE_SEG64,
  47         RANGE_SEG_GAP,
  48         RANGE_SEG_NUM_TYPES,
  49 } range_seg_type_t;
  50
  51 /*
  52  * Note: the range_tree may not be accessed concurrently; consumers
  53  * must provide external locking if required.
  54  */
  55 typedef struct range_tree {
  56         zfs_btree_t     rt_root;        /* offset-ordered segment b-tree */
  57         uint64_t        rt_space;       /* sum of all segments in the map */
  58         range_seg_type_t rt_type;       /* type of range_seg_t in use */
  59         /*
  60          * All data that is stored in the range tree must have a start higher
  61          * than or equal to rt_start, and all sizes and offsets must be
  62          * multiples of 1 << rt_shift.
  63          */
  64         uint8_t         rt_shift;
  65         uint64_t        rt_start;
  66         const range_tree_ops_t *rt_ops;
  67         void            *rt_arg;
  68         uint64_t        rt_gap;         /* allowable inter-segment gap */
  69
  70         /*
  71          * The rt_histogram maintains a histogram of ranges. Each bucket,
  72          * rt_histogram[i], contains the number of ranges whose size is:
  73          * 2^i <= size of range in bytes < 2^(i+1)
  74          */
  75         uint64_t        rt_histogram[RANGE_TREE_HISTOGRAM_SIZE];
  76 } range_tree_t;
  77
  78 typedef struct range_seg32 {
  79         uint32_t        rs_start;       /* starting offset of this segment */
  80         uint32_t        rs_end;         /* ending offset (non-inclusive) */
  81 } range_seg32_t;
  82
  83 /*
  84  * Extremely large metaslabs, vdev-wide trees, and dnode-wide trees may
  85  * require 64-bit integers for ranges.
  86  */
  87 typedef struct range_seg64 {
  88         uint64_t        rs_start;       /* starting offset of this segment */
  89         uint64_t        rs_end;         /* ending offset (non-inclusive) */
  90 } range_seg64_t;
  91
  92 typedef struct range_seg_gap {
  93         uint64_t        rs_start;       /* starting offset of this segment */
  94         uint64_t        rs_end;         /* ending offset (non-inclusive) */
  95         uint64_t        rs_fill;        /* actual fill if gap mode is on */
  96 } range_seg_gap_t;
  97
  98 /*
  99  * This type needs to be the largest of the range segs, since it will be stack
 100  * allocated and then cast the actual type to do tree operations.
 101  */
 102 typedef range_seg_gap_t range_seg_max_t;
 103
 104 /*
 105  * This is just for clarity of code purposes, so we can make it clear that a
 106  * pointer is to a range seg of some type; when we need to do the actual math,
 107  * we'll figure out the real type.
 108  */
 109 typedef void range_seg_t;
 110
 111 struct range_tree_ops {
 112         void    (*rtop_create)(range_tree_t *rt, void *arg);
 113         void    (*rtop_destroy)(range_tree_t *rt, void *arg);
 114         void    (*rtop_add)(range_tree_t *rt, void *rs, void *arg);
 115         void    (*rtop_remove)(range_tree_t *rt, void *rs, void *arg);
 116         void    (*rtop_vacate)(range_tree_t *rt, void *arg);
 117 };
 118
 119 static inline uint64_t
 120 rs_get_start_raw(const range_seg_t *rs, const range_tree_t *rt)
 121 {
 122         ASSERT3U(rt->rt_type, <=, RANGE_SEG_NUM_TYPES);
 123         switch (rt->rt_type) {
 124         case RANGE_SEG32:
 125                 return (((const range_seg32_t *)rs)->rs_start);
 126         case RANGE_SEG64:
 127                 return (((const range_seg64_t *)rs)->rs_start);
 128         case RANGE_SEG_GAP:
 129                 return (((const range_seg_gap_t *)rs)->rs_start);
 130         default:
 131                 VERIFY(0);
 132                 return (0);
 133         }
 134 }
 135
 136 static inline uint64_t
 137 rs_get_end_raw(const range_seg_t *rs, const range_tree_t *rt)
 138 {
 139         ASSERT3U(rt->rt_type, <=, RANGE_SEG_NUM_TYPES);
 140         switch (rt->rt_type) {
 141         case RANGE_SEG32:
 142                 return (((const range_seg32_t *)rs)->rs_end);
 143         case RANGE_SEG64:
 144                 return (((const range_seg64_t *)rs)->rs_end);
 145         case RANGE_SEG_GAP:
 146                 return (((const range_seg_gap_t *)rs)->rs_end);
 147         default:
 148                 VERIFY(0);
 149                 return (0);
 150         }
 151 }
 152
 153 static inline uint64_t
 154 rs_get_fill_raw(const range_seg_t *rs, const range_tree_t *rt)
 155 {
 156         ASSERT3U(rt->rt_type, <=, RANGE_SEG_NUM_TYPES);
 157         switch (rt->rt_type) {
 158         case RANGE_SEG32: {
 159                 const range_seg32_t *r32 = (const range_seg32_t *)rs;
 160                 return (r32->rs_end - r32->rs_start);
 161         }
 162         case RANGE_SEG64: {
 163                 const range_seg64_t *r64 = (const range_seg64_t *)rs;
 164                 return (r64->rs_end - r64->rs_start);
 165         }
 166         case RANGE_SEG_GAP:
 167                 return (((const range_seg_gap_t *)rs)->rs_fill);
 168         default:
 169                 VERIFY(0);
 170                 return (0);
 171         }
 172
 173 }
 174
 175 static inline uint64_t
 176 rs_get_start(const range_seg_t *rs, const range_tree_t *rt)
 177 {
 178         return ((rs_get_start_raw(rs, rt) << rt->rt_shift) + rt->rt_start);
 179 }
 180
 181 static inline uint64_t
 182 rs_get_end(const range_seg_t *rs, const range_tree_t *rt)
 183 {
 184         return ((rs_get_end_raw(rs, rt) << rt->rt_shift) + rt->rt_start);
 185 }
 186
 187 static inline uint64_t
 188 rs_get_fill(const range_seg_t *rs, const range_tree_t *rt)
 189 {
 190         return (rs_get_fill_raw(rs, rt) << rt->rt_shift);
 191 }
 192
 193 static inline void
 194 rs_set_start_raw(range_seg_t *rs, range_tree_t *rt, uint64_t start)
 195 {
 196         ASSERT3U(rt->rt_type, <=, RANGE_SEG_NUM_TYPES);
 197         switch (rt->rt_type) {
 198         case RANGE_SEG32:
 199                 ASSERT3U(start, <=, UINT32_MAX);
 200                 ((range_seg32_t *)rs)->rs_start = (uint32_t)start;
 201                 break;
 202         case RANGE_SEG64:
 203                 ((range_seg64_t *)rs)->rs_start = start;
 204                 break;
 205         case RANGE_SEG_GAP:
 206                 ((range_seg_gap_t *)rs)->rs_start = start;
 207                 break;
 208         default:
 209                 VERIFY(0);
 210         }
 211 }
 212
 213 static inline void
 214 rs_set_end_raw(range_seg_t *rs, range_tree_t *rt, uint64_t end)
 215 {
 216         ASSERT3U(rt->rt_type, <=, RANGE_SEG_NUM_TYPES);
 217         switch (rt->rt_type) {
 218         case RANGE_SEG32:
 219                 ASSERT3U(end, <=, UINT32_MAX);
 220                 ((range_seg32_t *)rs)->rs_end = (uint32_t)end;
 221                 break;
 222         case RANGE_SEG64:
 223                 ((range_seg64_t *)rs)->rs_end = end;
 224                 break;
 225         case RANGE_SEG_GAP:
 226                 ((range_seg_gap_t *)rs)->rs_end = end;
 227                 break;
 228         default:
 229                 VERIFY(0);
 230         }
 231 }
 232
 233 static inline void
 234 rs_set_fill_raw(range_seg_t *rs, range_tree_t *rt, uint64_t fill)
 235 {
 236         ASSERT3U(rt->rt_type, <=, RANGE_SEG_NUM_TYPES);
 237         switch (rt->rt_type) {
 238         case RANGE_SEG32:
 239                 /* fall through */
 240         case RANGE_SEG64:
 241                 ASSERT3U(fill, ==, rs_get_end_raw(rs, rt) - rs_get_start_raw(rs,
 242                     rt));
 243                 break;
 244         case RANGE_SEG_GAP:
 245                 ((range_seg_gap_t *)rs)->rs_fill = fill;
 246                 break;
 247         default:
 248                 VERIFY(0);
 249         }
 250 }
 251
 252 static inline void
 253 rs_set_start(range_seg_t *rs, range_tree_t *rt, uint64_t start)
 254 {
 255         ASSERT3U(start, >=, rt->rt_start);
 256         ASSERT(IS_P2ALIGNED(start, 1ULL << rt->rt_shift));
 257         rs_set_start_raw(rs, rt, (start - rt->rt_start) >> rt->rt_shift);
 258 }
 259
 260 static inline void
 261 rs_set_end(range_seg_t *rs, range_tree_t *rt, uint64_t end)
 262 {
 263         ASSERT3U(end, >=, rt->rt_start);
 264         ASSERT(IS_P2ALIGNED(end, 1ULL << rt->rt_shift));
 265         rs_set_end_raw(rs, rt, (end - rt->rt_start) >> rt->rt_shift);
 266 }
 267
 268 static inline void
 269 rs_set_fill(range_seg_t *rs, range_tree_t *rt, uint64_t fill)
 270 {
 271         ASSERT(IS_P2ALIGNED(fill, 1ULL << rt->rt_shift));
 272         rs_set_fill_raw(rs, rt, fill >> rt->rt_shift);
 273 }
 274
 275 typedef void range_tree_func_t(void *arg, uint64_t start, uint64_t size);
 276
 277 range_tree_t *range_tree_create_gap(const range_tree_ops_t *ops,
 278     range_seg_type_t type, void *arg, uint64_t start, uint64_t shift,
 279     uint64_t gap);
 280 range_tree_t *range_tree_create(const range_tree_ops_t *ops,
 281     range_seg_type_t type, void *arg, uint64_t start, uint64_t shift);
 282 void range_tree_destroy(range_tree_t *rt);
 283 boolean_t range_tree_contains(range_tree_t *rt, uint64_t start, uint64_t size);
 284 range_seg_t *range_tree_find(range_tree_t *rt, uint64_t start, uint64_t size);
 285 boolean_t range_tree_find_in(range_tree_t *rt, uint64_t start, uint64_t size,
 286     uint64_t *ostart, uint64_t *osize);
 287 void range_tree_verify_not_present(range_tree_t *rt,
 288     uint64_t start, uint64_t size);
 289 void range_tree_resize_segment(range_tree_t *rt, range_seg_t *rs,
 290     uint64_t newstart, uint64_t newsize);
 291 uint64_t range_tree_space(range_tree_t *rt);
 292 uint64_t range_tree_numsegs(range_tree_t *rt);
 293 boolean_t range_tree_is_empty(range_tree_t *rt);
 294 void range_tree_swap(range_tree_t **rtsrc, range_tree_t **rtdst);
 295 void range_tree_stat_verify(range_tree_t *rt);
 296 uint64_t range_tree_min(range_tree_t *rt);
 297 uint64_t range_tree_max(range_tree_t *rt);
 298 uint64_t range_tree_span(range_tree_t *rt);
 299
 300 void range_tree_add(void *arg, uint64_t start, uint64_t size);
 301 void range_tree_remove(void *arg, uint64_t start, uint64_t size);
 302 void range_tree_remove_fill(range_tree_t *rt, uint64_t start, uint64_t size);
 303 void range_tree_adjust_fill(range_tree_t *rt, range_seg_t *rs, int64_t delta);
 304 void range_tree_clear(range_tree_t *rt, uint64_t start, uint64_t size);
 305
 306 void range_tree_vacate(range_tree_t *rt, range_tree_func_t *func, void *arg);
 307 void range_tree_walk(range_tree_t *rt, range_tree_func_t *func, void *arg);
 308 range_seg_t *range_tree_first(range_tree_t *rt);
 309
 310 void range_tree_remove_xor_add_segment(uint64_t start, uint64_t end,
 311     range_tree_t *removefrom, range_tree_t *addto);
 312 void range_tree_remove_xor_add(range_tree_t *rt, range_tree_t *removefrom,
 313     range_tree_t *addto);
 314
 315 #ifdef  __cplusplus
 316 }
 317 #endif
 318
 319 #endif  /* _SYS_RANGE_TREE_H */